TIKTOK SHOP · JUN 13, 2026 · 8 MIN

How to Find Undiscovered TikTok Shop Creators with Claude Code and the Cruva MCP

Your competitors have already enrolled the creators on the top 100 list. The leverage is finding the dark horse creators they have not noticed yet, the ones growing 500 to 1,000 percent week over week off a small base. With Claude Code wired into the Cruva API, you can pull them out of any competitor's affiliate roster in a single run. Here is the exact workflow.

The top 100 TikTok Shop creators are saturated. Every brand in the category is already in their DMs, and the marginal lift you get from signing one more is small. The leverage in 2026 is finding the dark horse creators your competitors are nurturing but have not noticed yet, the ones growing 500 to 1,000 percent week over week off a small base. With Claude Code wired into the Cruva API, you can pull them out of any competitor's affiliate roster in a single run.

The short answer, for anyone who landed here from a search: in Claude Code, define what "fast growing" means for you (we use 50 percent or more week over week), name a small set of competitor brands, and prompt Claude to fetch their products from the Cruva API, pull the creators attached to those products, remove anyone in the top 100, and rank what remains by growth rate. Claude writes and runs the script, returns the result as an HTML table, and you take it from there. The whole run is under five minutes.

The rest of this piece walks through the recruiting logic behind the workflow, the prompt structure, and a worked example from a real run against an electrolyte category.

Why Dark Horse Creator Recruiting Is the Highest-Leverage Move

Creator recruiting on TikTok Shop is a relationship game with the same shape as sports recruiting. The blue-chip prospects (the top 100 creators) get offers from everyone, sign with whoever pays the most, and stay only as long as the offer holds. The leverage is at the tier below: creators who are clearly on a steep growth curve, are still small enough that the relationship cost is low, and are not yet being chased by a dozen brands.

If you find a creator at 1,000 percent week-over-week growth off a $100 base, you are catching them at a moment where:

That is the moment to form the relationship. Two months later that same creator is at $20,000 a week and signing a contract you could not afford to pitch in the first place.

Step 1: Define "Fast Growing" Before You Prompt

This is the prompt move people skip. Tell Claude exactly what "fast growing" means for your program, as a number. We use 50 percent or more week over week as the floor. Above that, the creator is on a real trajectory. Below it, you are looking at week-to-week noise.

The reason a number matters: without one, Claude returns "creators with strong growth," which sounds useful and is not. With a numeric rule, every name on the output meets a defined criterion you can defend to a brand lead.

Step 2: Exclude the Top 100

The second rule, and the one that turns this from "creator list" into "dark horse list": tell Claude to remove anyone in the top 100 on the marketplace. This is the exclusion that creates the edge.

Without it, the highest-growth creators in your output are the ones already on a viral run, already being chased by every brand in the category, already represented. With the exclusion, what is left is the under-the-radar tier: creators with the trajectory of a top-100 prospect, three months before they get there.

Step 3: Pick a Competitor Set to Poach From

You are not pulling creators from the open marketplace. You are pulling them from the affiliate rosters of competitor brands you have already decided are interesting. If you ran the competitive analysis workflow first, you already know which competitors to point at, the ones with strong recruitment but who you suspect are not nurturing past enrollment.

In the electrolyte category, the natural set is Bloom, Liquid IV, and a handful of category peers. Pick three to five brands. Going wider than that dilutes the signal; going narrower than that misses creators who post for multiple brands in the category.

Step 4: Let Claude Code Write and Run the Script

This is where Claude Code earns its keep. You do not write the script. You describe the rule, Claude writes a script that hits the Cruva API, fetches each competitor's products, pulls the creators attached to those products, computes the week-over-week growth rate on each, drops the top 100 from the marketplace, and ranks what remains.

A condensed version of the prompt, ready to adapt:

Using the Cruva API, find dark horse creators in [CATEGORY]. A dark horse is a creator with 50 percent or higher week-over-week GMV growth who is not in the top 100 on the marketplace. Pull the product rosters for [COMPETITOR BRANDS]. For each product, fetch the affiliate creators and their last two weeks of GMV. Compute week-over-week growth. Drop top 100 creators. Rank the remainder by growth rate. Output as an HTML table I can scan with creator handle, brand they are currently posting for, last week GMV, this week GMV, and growth rate.

The HTML output matters. A JSON dump is unreadable in a single pass. An HTML table is a scorecard you can scan in 60 seconds and decide who to look at first.

Step 5: Spend the Real Time on the Creators, Not the Pull

The pull is the cheap part. The expensive part is the next move: spending 15 to 20 minutes on each shortlisted creator's profile, watching their recent content, getting a read on their voice and audience, and putting together an offer that lands.

The reason the workflow is set up this way is to push the expensive minutes onto the part that actually wins the creator. Most brand teams spend hours building the list and minutes on the pitch. The workflow flips it: minutes on the list, hours on the pitch.

A Worked Example: An Electrolyte Category Run

I ran this workflow against the electrolyte category, with Bloom and Liquid IV as the primary competitor set. The Cruva pull returned the products, the affiliate rosters came back attached, the top 100 exclusion stripped the obvious names, and the growth ranking surfaced the dark horses.

The leaderboard at the top of the HTML output:

The absolute numbers are not the point. The first creator is still doing $1,400 a week, which is invisible on Bloom's standard affiliate dashboard. The point is the trajectory. At 1,130 percent week over week, that creator is two or three weeks from $10,000 weekly GMV. Bloom has clearly enrolled them, but at $96 a week of GMV, nobody on the Bloom team is watching this creator yet. That is the window.

Two of those three creators became calls within the week. Both signed with the brand we were running the workflow for, on terms that would have been impossible to negotiate ninety days later.

Why Claude Code, Not Just an MCP Chat Connector

A chat connector is the right tool when the work is "ask a question, get an answer." This is not that. This is a small, repeatable data pipeline: hit the API, transform the result, apply two filter rules, sort, render as HTML. Claude Code is built for exactly this shape: it writes the script, runs it against your terminal, iterates if the schema is not quite what was expected, and saves the result to disk.

If you are running the weekly reporting workflow, the MCP chat connector is the right surface. For the recruiting workflow, Claude Code is the surface, because you want the script you can rerun every Monday morning without having to re-prompt.

Where This Sits in the Affiliate Stack

The competitive analysis tells you which competitors to learn from. The dark horse pull tells you which creators to recruit from those competitors. The AI-generated brief workflow is what you hand the creator after they sign, so the content they post for you actually converts. And the weekly reporting workflow is how you measure whether any of it worked.

Together, those four pieces are a fully connected affiliate operation, with the same underlying Cruva data layer powering each one. The dark horse pull is the recruiting node, and it is the one with the most leverage per minute of operator time.

If you run a TikTok Shop program and want help wiring this workflow into your own stack, get in touch.

Frequently Asked Questions

What counts as a dark horse TikTok Shop creator?

A creator whose week-over-week GMV growth is high (we use 50 percent or more as the floor) but who is not in the top 100 on the marketplace. The top 100 are already saturated with brand offers. The dark horses are off everyone's radar and still cheap to sign.

Why exclude the top 100 creators from the search?

Because everyone is already in their DMs. By the time a creator is in the top 100, the relationship cost goes up, the exclusivity terms get harder, and the marginal lift you get from adding them is small. The dark horses are where the recruiting leverage actually lives.

Do I need to write code to run this?

You write a prompt in Claude Code that describes the rule (growth rate, top 100 exclusion, competitor set) and Claude writes and runs the script against the Cruva API. The skill is defining the recruiting rule, not the engineering.

How is this different from the competitive analysis workflow?

The competitive analysis benchmarks your program against a competitor set on recruitment, retention, conversion, and GMV. This workflow is downstream of that: once you know which competitors to learn from, this is how you poach the creators they are nurturing but have not noticed yet.

Does this work with tools other than Cruva?

Yes. Any TikTok Shop data source with API or MCP access to creator-level growth data works. The defining rule (high growth, exclude top 100) is tool-agnostic; the metric framework is what carries.

How often should I run this?

Weekly. The growth windows on TikTok Shop close fast. A creator at 1,000 percent week-over-week growth this Monday is a different conversation by next Monday, when their numbers are visible to everyone. We run the pull every Monday morning and the outreach the same afternoon.